ICGB key terms

A structured collection of customer needs, expectations, and complaints gathered through interviews, surveys, and complaint logs, used to define what a process must deliver before any measurement or improvement work begins.

A measurable characteristic that translates a broad customer need into a specific, verifiable requirement, such as delivery within 48 hours or a defect rate below one percent.

A short authorizing document stating the problem, goal, scope, team, timeline, and expected benefit of a Six Sigma project, signed by a sponsor who can remove barriers.

A factual description of the gap between current and desired performance, including where, when, and how much, written without stating a cause or a solution.

A specific, measurable, time-bound target that the project must reach, expressed in the same metric and units as the problem statement so progress can be judged objectively.

A high-level map showing Suppliers, Inputs, Process steps, Outputs, and Customers, used at project definition to agree on scope and boundaries without detailing every task.

A senior manager who sponsors a Six Sigma project, secures resources, resolves cross-functional conflicts, and translates project results into organizational decisions.

The hierarchy of Six Sigma practitioners, typically Yellow, Green, Black, and Master Black Belt, distinguished by training depth, tools mastered, and the size of projects led.

The five-phase improvement cycle of Define, Measure, Analyze, Improve, and Control, applied to existing processes whose performance must be raised to a new level.

A design cycle of Define, Measure, Analyze, Design, and Verify, used when no process exists yet or an existing one cannot meet requirements and must be replaced.

A precise description of how a characteristic is measured, including the instrument, procedure, and decision criteria, so that different people obtain the same result.

A study that quantifies how much variation in recorded data comes from the gauge, the appraisers, and the procedure rather than from the process itself.

A repeatability and reproducibility study that separates variation when one person measures the same item twice from variation between different appraisers measuring the same items.

An MSA method for pass/fail or categorical judgments that checks whether appraisers agree with themselves, with each other, and with a known reference standard.

The average difference between measured values and a reference or true value across the operating range, indicating a systematic offset rather than random scatter.

The consistency of a measurement system's bias across the range of values being measured, so that accuracy does not deteriorate at the high or low end.

A chance-corrected index of agreement between raters or between a rater and a standard, where values near one indicate agreement beyond what random assignment would produce.

Indices comparing specification width with process spread (Cp) and with the distance from process mean to the nearer specification limit (Cpk), assuming stable, roughly normal data.

Capability-style indices computed from the overall standard deviation of all data, including between-subgroup shifts, so they describe what was actually produced rather than potential capability.

The number of defects found divided by the total number of defect opportunities, multiplied by one million, giving a unit-free quality level comparable across processes.

The number of standard deviations between the process mean and the nearest specification limit, commonly converted from DPMO to express process quality on a single scale.

The probability that a unit passes every step of a multi-step process without any defect, obtained by multiplying the first-pass yields of the individual steps.

The proportion of units that meet specification the first time through a step or process, without rework, retest, or repair being counted as success.

The classification of measurements as continuous when they can take any value on a scale, or discrete when they are counts or categories, which determines which statistical tools apply.

A documented rule for how many observations to take, how to select them, and how often, chosen so the data represent the process and support the intended statistical test.

Separating data into layers such as shift, machine, operator, or supplier to reveal patterns that are hidden when all observations are pooled together.

A bar chart of categories ordered by frequency or cost, paired with a cumulative line, used to identify the few contributors that account for most of the problem.

A branching diagram that organizes possible causes of a problem under categories such as people, methods, machines, materials, measurement, and environment, also called a fishbone.

An iterative questioning technique that asks why a problem occurred repeatedly until the underlying systemic cause is reached rather than a symptom or an individual's mistake.

The step of confirming a suspected cause by turning it on and off, checking data, or observing the effect change, before committing resources to a countermeasure.

A statistical procedure that compares observed data with what a stated null hypothesis predicts, yielding a p-value used to decide whether an effect is likely real.

The probability of observing data at least as extreme as the sample results if the null hypothesis were true; it is not the probability that the null hypothesis is true.

Type I error rejects a true null hypothesis, a false alarm, while Type II error fails to reject a false null hypothesis, a missed effect; power is one minus Type II error.

A range of plausible values for a population parameter, computed so that a stated percentage of such intervals from repeated sampling would contain the true value.

Analysis of variance, a test comparing the means of three or more groups by judging whether between-group variation is large relative to within-group variation.

A modeling technique that estimates how a response variable changes with one or more predictors, quantifying both the strength and the form of the relationship.

A measure of the strength and direction of linear association between two variables, which by itself does not establish that one variable causes the other.

A planned set of runs in which factors are varied together according to a statistical structure, so main effects and interactions can be estimated efficiently.

An experiment that runs only a chosen subset of the full factorial combinations, trading the ability to estimate some high-order effects for fewer runs and lower cost.

A main effect is the average change in response when one factor moves between levels; an interaction means the effect of one factor depends on the level of another.

A device or design feature that prevents an error from being made or makes an incorrect action impossible to complete, rather than relying on vigilance or inspection.

A documented, current best sequence of tasks with defined cycle times, takt time, and work-in-process, forming the baseline against which deviations are detected.

A living document listing each process characteristic, its specification, the measurement method, sample size, frequency, and the reaction plan when results fall outside limits.

A time-ordered plot of a process statistic with statistically derived limits that distinguish common-cause variation from signals of special-cause change.

Common cause is the routine, inherent variation of a stable process; special cause is unusual variation from an identifiable source that a control chart is designed to flag.

A predefined sequence telling operators whom to notify and what containment or adjustment to perform when a control chart signals an out-of-control condition.

A focused, rapid improvement event in which the people who do the work redesign a process together, typically over several consecutive days, with immediate implementation.

The Plan-Do-Check-Act cycle of iterative learning: plan a change, trial it, compare results with expectations, and standardize or adjust before repeating.

Calculating how many observations are needed to detect an effect of practical size with chosen confidence and power, before data collection begins.

A step-by-step diagram of how work actually flows, including decisions, handoffs, and rework loops, used to expose delays and non-value-adding steps.